An Integrative Segmentation Framework for Cell Nucleus of Fluorescence Microscopy.

Area of Science:

• Cell Biology
• Bioimaging
• Computational Biology

Background:

• Accurate nucleus segmentation is vital for quantitative cell biology.
• Existing methods struggle with noisy images and clumped nuclei.
• Cascaded U-Net architectures show promise in medical image segmentation.

Purpose of the Study:

• To develop a novel, accurate, and lightweight framework for nucleus segmentation in fluorescence microscopy.
• To address limitations of current methods in handling challenging imaging conditions.
• To improve the analysis of nucleus morphology.

Main Methods:

• Proposed the Attention-enhanced Simplified W-Net (ASW-Net) framework.
• Implemented a cascade-like structure with between-net connections.
• Evaluated performance on the BBBC039 testing set.

Main Results:

• ASW-Net achieved a high aggregated Jaccard index of 0.90 on the BBBC039 dataset.
• The proposed framework outperformed existing state-of-the-art methods.
• Deep feature visualization confirmed the network's effectiveness.

Conclusions:

• ASW-Net offers remarkable segmentation performance for challenging microscopy images.
• The lightweight and open-source nature of ASW-Net facilitates broader adoption.
• This method enhances quantitative analysis in cell biology.